diff options
author | Brecht Van Lommel <brechtvanlommel@pandora.be> | 2012-05-28 23:21:13 +0400 |
---|---|---|
committer | Brecht Van Lommel <brechtvanlommel@pandora.be> | 2012-05-28 23:21:13 +0400 |
commit | 131de4352b9e8f3e4ce77d2d9dc145f5db461aed (patch) | |
tree | c796dc4d9697a1122383dea92aab31493e9c0e1a /intern/cycles/device | |
parent | c2be2fd4083845a381c2462a65dc0856dc6efb8e (diff) |
Cycles: fixes to make CUDA 4.2 work, compiling gave errors in shadows and
other places, was mainly due to instancing not working, but also found
issues in procedural textures.
The problem was with --use_fast_math, this seems to now have way lower
precision for some operations. Disabled this flag and selectively use
fast math functions. Did not find performance regression on GTX 460 after
doing this.
Diffstat (limited to 'intern/cycles/device')
-rw-r--r-- | intern/cycles/device/device_cuda.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp index 23ad5e5ff92..491a63a7cf2 100644 --- a/intern/cycles/device/device_cuda.cpp +++ b/intern/cycles/device/device_cuda.cpp @@ -259,7 +259,7 @@ public: path_create_directories(cubin); - string command = string_printf("\"%s\" -arch=sm_%d%d -m%d --cubin \"%s\" --use_fast_math " + string command = string_printf("\"%s\" -arch=sm_%d%d -m%d --cubin \"%s\" " "-o \"%s\" --ptxas-options=\"-v\" --maxrregcount=%d --opencc-options -OPT:Olimit=0 -I\"%s\" -DNVCC", nvcc.c_str(), major, minor, machine, kernel.c_str(), cubin.c_str(), maxreg, include.c_str()); |